DeepSeek released an updated V4.1-Flash model with higher scores and shorter outputs

DeepSeek released an updated version of its V4.1-Flash AI model on Saturday that the company said finishes tasks more quickly and writes more concise responses than its predecessor. The model, called DeepSeek-V4.1-Flash-1017, is live on DeepSeek's website, app and API.

The API calling method and pricing are unchanged, and users reach the new model by using the name deepseek-flash, the company said. DeepSeek also released the model as open source on Hugging Face and ModelScope.

DeepSeek's change log lists a score of 42.6 on Humanity's Last Exam without tools, up from 36.8 for the September release of V4.1-Flash. On Terminal-Bench 4.0, the new model scored 42.3, compared with 31.2 for the September version. ProgramBench rose to 29.1 from 20.3, and DeepSWE v1.1 rose to 77.0 from 74.2. DeepSeek said it tested Terminal-Bench 4.0 with the Claude Code framework.

DeepSeek said the model uses reinforcement learning with self-feedback to improve end-to-end project development, from game creation to 3D modeling. The company said it optimized the model for DeepSeek Harness, Tencent CodeBuddy and WorkBuddy, and it recommends upgrading to DeepSeek Harness 0.2.1-rc.1 or later.

DeepSeek opened its Search & Fetch APIs to everyone as part of the update. Each endpoint costs $0.75 per 1,000 calls. Calls made through server-side tools in DeepSeek Harness, Claude Code and Codex are free for DeepSeek API users until billing starts at 16:00 UTC on Oct. 31, 2026.

DeepSeek called the release an incremental revision of V4.1-Flash, according to Bloomberg. The launch adds to competition between DeepSeek and rivals including Anthropic, OpenAI, Moonshot Kimi and Z.ai, according to Bloomberg. Bloomberg also described a rivalry pitting Chinese open-weight systems against American closed models, which it said lead in performance but carry much steeper prices.

DeepSeek's servers were unreachable for about five hours last week, according to Bloomberg. Some experts blamed insufficient computing capacity for the disruption, according to Bloomberg.

DeepSeek released the original V4.1-Flash on Sept. 10 and described it as the smallest model in its new architecture family, with native multimodal visual understanding. It cut API prices with that release.

DeepSeek is also raising money. It was on the verge of raising at least 80 billion yuan ($12 billion) as of early October. The round values the company at a minimum of 500 billion yuan, with Tencent Holdings and Contemporary Amperex Technology Co. among the top investors.